Package ubic.basecode.math.linearmodels

Class LinearModelSummary

  • All Implemented Interfaces:
    Serializable
    public class LinearModelSummary
extends Object
implements Serializable
    Represents the results of a linear model analysis from R. Both the summary.lm and anova objects are represented. FIXME make this have the capabilities of the "fit" object, rather than just the summary.
    Author:
    paul
    See Also:
    Serialized Form

    • Constructor Detail

      • LinearModelSummary

        public LinearModelSummary​(org.rosuda.REngine.REXP summaryLm,
                          org.rosuda.REngine.REXP anova,
                          String[] factorNames)
        Construct from the result of evaluation of an R call. should be deprecated because we're not using R integration
        Parameters:
        summaryLm -
        anova -
        factorNames - as referred to in the model. Used as keys to keep track of coefficients etc.

      • LinearModelSummary

        public LinearModelSummary​(String key)
        Construct an empty summary. Use for model fits that fail due to 0 degrees of freedom, etc.
        Parameters:
        key - identifier

      • LinearModelSummary

        public LinearModelSummary​(String k,
                          Double[] coefficients,
                          Double[] residuals,
                          List<String> terms,
                          DoubleMatrix<String,​String> contrastCoefficients,
                          Double[] effects,
                          Double[] stdevUnscaled,
                          double rsquared,
                          double adjRsquared,
                          double fstat,
                          Integer ndof,
                          Integer ddof,
                          GenericAnovaResult anovaResult,
                          double sigma,
                          boolean isShrunken)
        Parameters:
        k - optional identifier
        coefficients -
        residuals -
        terms -
        contrastCoefficients -
        effects - AKA Qty
        rsquared -
        adjRsquared -
        fstat -
        ndof -
        ddof -
        anovaResult -
        sigma -
        isShrunken -

      • LinearModelSummary

        protected LinearModelSummary()
        Construct an empty summary. Use for model fits that fail due to 0 residual degrees of freedom, etc.

    • Method Detail

      • getAdjRSquared

        public Double getAdjRSquared()
        Returns:
        the adjRSquared

      • getAnova

        public GenericAnovaResult getAnova()
        Returns:
        may be null if ANOVA was not run.

      • getCoefficients

        public Double[] getCoefficients()

      • getContrastCoefficients

        public DoubleMatrix<String,​String> getContrastCoefficients()
        Returns:
        The contrast coefficients and associated statistics for all tested contrasts.

        Row names are the contrasts, for example for a model with one factor "f" with two levels "a" and "b": {"(Intercept)", "fb"}. columns are always {"Estimate" ,"Std. Error", "t value", "Pr(>|t|)"}

      • getContrastCoefficients

        public Map<String,​Double> getContrastCoefficients​(String factorName)
        Parameters:
        factorName -
        Returns:

      • getContrastCoefficientStderr

        public Map<String,​Double> getContrastCoefficientStderr​(String factorName)
        For the requested factor, return the standard errors associated with the contrast coefficient estimates.
        Parameters:
        factorName -
        Returns:

      • getContrastPValues

        public Map<String,​Double> getContrastPValues​(String factorName)
        Parameters:
        factorName -
        Returns:
        Map of pvalues for the given factor. For continuous factors or factors with only one level, there will be just one value. For factors with N>2 levels, there will be N-1 values, one for each contrast (since we compute treatment contrasts to the baseline)

      • getContrastTStats

        public Map<String,​Double> getContrastTStats​(String factorName)
        Parameters:
        factorName -
        Returns:
        Map of T statistics for the given factor. For continuous factors or factors with only one level, there will be just one value. For factors with N>2 levels, there will be N-1 values, one for each contrast (since we compute treatment contrasts to the baseline)

      • getEffects

        public Double[] getEffects()
        Returns:

      • getF

        public Double getF()
        Returns:
        F statistic for overall model fit.

      • getFactorNames

        public List<String> getFactorNames()
        Returns:
        the factorNames

      • getFactorValueNames

        public List<String> getFactorValueNames​(String factorName)
        Return the factor names in the order they are stored here. Pvalues and T statistics for this factor are in the same order, but the 'baseline' must be accounted for.
        Parameters:
        factorName -
        Returns:

      • getFormula

        public String getFormula()
        Returns:
        the formula

      • getInteractionEffectP

        public Double getInteractionEffectP​(String... fnames)
        Parameters:
        fnames - names of the factors
        Returns:
        See Also:
        ubic.basecode.math.linearmodels.GenericAnovaResult#getInteractionEffectP(java.lang.String)

      • getInterceptCoeff

        public Double getInterceptCoeff()
        Returns:

      • getInterceptP

        public Double getInterceptP()
        Returns:

      • getInterceptT

        public Double getInterceptT()
        Returns:

      • getKey

        public String getKey()

      • getNumeratorDof

        public Integer getNumeratorDof()
        Returns:

      • getP

        public Double getP()
        Overall p value for F stat of model fit (upper tail probability)
        Returns:
        value or NaN if it can't be computed for some reason

      • getPriorDof

        public Double getPriorDof()
        Returns:
        the priorDof

      • getResidualDof

        public Integer getResidualDof()
        Returns:

      • getResiduals

        public Double[] getResiduals()
        Returns:
        the residuals

      • getRSquared

        public Double getRSquared()
        Returns:
        the rSquared

      • getSigma

        public Double getSigma()
        Residual standard deviation
        Returns:

      • getStdevUnscaled

        public Double[] getStdevUnscaled()
        Unscaled standard deviations for the coefficient estimators in same order as coefficients. The standard errors are given by stdev.unscaled * sigma (a la limma)

      • isBaseline

        public boolean isBaseline​(String factorValueName)

      • isShrunken

        public boolean isShrunken()
        Whether this is the result of emprical bayes shrinkage of variance estimates
        Returns:

      • setAnova

        public void setAnova​(GenericAnovaResult genericAnovaResult)
        Parameters:
        genericAnovaResult -

      • setKey

        public void setKey​(String key)

      • setPriorDof

        public void setPriorDof​(Double priorDof)
        Parameters:
        priorDof - the priorDof to set